Multi-element comparative analysis of ship assembly construction digital upgrade architecture construction method

Abstract

The application provides a ship assembly construction digital upgrading architecture construction method for element comparison analysis, and comprises the following steps: S1: ship assembly construction digital business architecture multi-element comparison analysis; S2: ship assembly construction digital business element library construction; S3: ship assembly construction digital upgrading business architecture construction; S4: ship assembly construction digital upgrading application architecture construction; S5: ship assembly construction digital upgrading data architecture construction; and S6: ship assembly construction digital upgrading technology architecture construction. The application can solve the technical problems of the prior art, such as incomplete architecture construction method, lack of quantitative analysis means, and lack of linkage mechanism between architecture levels, realizes the leap from qualitative description to quantitative calculation, and provides a systematic and quantifiable architecture construction scheme for the digital upgrading of ship assembly construction.

Description

Technical Field

[0001] This invention relates to the field of digital technology in ship assembly and construction, and in particular to a method for constructing a digital upgrade architecture for ship assembly and construction based on multi-factor comparative analysis. Background Technology

[0002] Ship assembly and construction is a typical complex discrete manufacturing process, involving the collaborative operations of tens of thousands of parts, hundreds of process nodes, and dozens of specialized departments. With the rapid development of new-generation information technologies such as the Industrial Internet of Things, big data analytics, and artificial intelligence, the digital transformation of ship assembly and construction has become a key approach to improving shipbuilding efficiency and quality. However, the digital upgrade of ship assembly and construction faces numerous technical challenges: the business architecture has many elements and complex relationships between them; the application architecture involves multiple fields and has a large data scale; the data architecture is multi-source and heterogeneous with a large data volume; the technical architecture needs to be deeply integrated with new-generation information technologies; and the information security system needs to cover a complex network environment with many types of equipment and numerous information systems.

[0003] In the field of digitalization of ship assembly and construction, existing technical solutions mainly focus on single-level digital improvements. Some existing solutions focus on the tracking and optimization of data service call links. This involves generating service node topology maps by acquiring full lifecycle datasets, performing anomaly detection and processing on task execution results, generating a link tracking feature matrix, and then generating service call link optimization strategies, emphasizing the technical implementation at the data service level. Other existing solutions focus on inspection management during ship assembly and construction. This involves generating management templates based on product structure, production processes, and technological requirements through template generation units, modifying and filling in templates through editing units, and achieving information sharing through display units, primarily addressing the digital management of inspection business processes. Furthermore, some existing solutions focus on optimizing the production plan for ship assembly. This involves determining critical links by setting the production cycle and sequence relationship of each section, finding the longest construction timeline as the critical link based on the end-time node alignment principle, and achieving digital management through iterative calculation methods for critical links. Still other existing solutions focus on comprehensive information integration, using calculations of coordination order, task completion index, and comprehensive information registration index to build system frameworks and provide early warnings.

[0004] The aforementioned existing technical solutions all suffer from the following shortcomings: First, they only involve single-level technical improvements in the digital upgrade of ship assembly and construction, lacking a complete and interconnected construction method encompassing business architecture, application architecture, data architecture, technical architecture, and information security systems. This results in each architectural layer being independent and lacking coordination, failing to form a unified digital upgrade architecture system. Second, in business architecture analysis, there is a lack of mathematical modeling methods for the hierarchical relationships between business elements and departments or business domains. The sorting and difference analysis of business elements rely on manual experience and judgment, making it impossible to quantitatively assess the differences in business architecture elements and upgrade needs, leading to unclear upgrade directions and a lack of data support for decision-making. Third, in application architecture construction, existing technical solutions lack modular decoupling methods based on microservice architecture and data interface design methods based on event-driven architecture. The high coupling between application modules makes it difficult to adapt to changes in business needs and the expansion of system scale. Interface design lacks standardization, resulting in low system integration efficiency. Fourth, the construction of data architecture lacks a multi-source heterogeneous data fusion method based on data quality assessment. The quality differences between different data sources cannot be quantitatively assessed, data fusion lacks a scientific basis, the negative impact of low-quality data on the fusion results cannot be effectively controlled, and data management lacks a standardized system. Summary of the Invention

[0005] This invention provides a method for constructing a digital upgrade architecture for ship assembly and construction based on multi-factor comparative analysis. Taking the entire lifecycle business process of ship assembly and construction as the research object, and using formal concept analysis theory, fuzzy clustering theory, grey relational analysis theory, graph neural network theory, data flow statistical methods, data envelopment analysis theory, and entropy weight method theory as mathematical foundations, it sequentially constructs a business architecture upgrade requirement vector, a business element library for digital upgrade of ship assembly and construction, a business architecture for digital upgrade of ship assembly and construction, an application architecture for digital upgrade of ship assembly and construction, a data architecture for digital upgrade of ship assembly and construction, and a technical architecture for digital upgrade of ship assembly and construction, ultimately forming a complete architecture system for digital upgrade of ship assembly and construction.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison and analysis includes the following steps: S1: Obtain current business architecture data and target business architecture data, define a standardized set of business domains, construct the current form background, current form concept grid, target form background and target form concept grid, and generate a business architecture upgrade requirement vector; S2: Based on the business architecture upgrade requirement vector, classify and merge business elements, establish the relationship between the merged business elements, and generate a business element library for the digital upgrade of ship assembly and construction. S3: Based on the business element library for the digital upgrade of ship assembly and construction, a business architecture diagram is constructed for each standardized business domain in the standardized business domain set, and the business architecture feature representation is extracted. A business architecture layer between standardized business domains is constructed, the cohesion of business elements within standardized business domains and the coupling of business elements between standardized business domains are optimized, and a business architecture for the digital upgrade of ship assembly and construction is generated. The business architecture for the digital upgrade of ship assembly and construction includes a business architecture hypergraph. S4: Map the hyperedges in the business architecture hypergraph to application function modules, decouple the application function modules and design the data interfaces between each module to generate the application architecture for digital upgrade of ship assembly and construction. S5: Based on the application architecture for digital upgrade of ship assembly and construction, data quality assessment is performed on multi-source heterogeneous data sources. Based on the data quality assessment results, multi-source heterogeneous data is integrated. The integrated data is managed using a layered architecture to generate a data architecture for digital upgrade of ship assembly and construction. S6: Based on the data architecture for digital upgrade of ship assembly and construction, and combined with the requirements of the business architecture and application architecture for digital upgrade of ship assembly and construction, identify technical requirements and select technologies to generate the technical architecture for digital upgrade of ship assembly and construction.

[0007] In summary, the present invention has at least the following beneficial effects: This invention constructs formal background and formal concept lattice between business architecture elements and standardized business domains through formal concept analysis theory, realizing mathematical modeling of business architecture elements and departmental affiliation relationships. This transforms business architecture analysis, which originally relied on human experience judgment, into a computable and quantifiable mathematical problem, significantly improving the objectivity and accuracy of business architecture analysis.

[0008] This invention uses fuzzy clustering to automatically merge business elements that need to be integrated, and uses grey relational analysis to calculate the grey relational degree between business elements and establish a quantitative relational matrix. This achieves automation and quantification of business element changes and relational establishment, overcoming the inefficiency and subjective defects of traditional methods that rely on manual sorting.

[0009] This invention uses graph attention networks to learn the representation of business architecture within a standardized business domain. It models business elements and their relationships within a department as graph-structured data and propagates them through graph neural networks to extract feature representations of the department's business architecture, thus achieving adaptive extraction and deep representation of business architecture features.

[0010] This invention objectively calculates the interaction strength between standardized business domains using data flow statistics, replacing the subjective scoring that relies on expert experience in traditional methods, and achieving objective quantification of the logical relationships between business domains.

[0011] This invention achieves quantitative decoupling of application functional modules through a microservice architecture, designs data interfaces between modules through an event-driven architecture, and achieves seamless integration with existing systems through an adapter module. This significantly reduces the coupling between modules and improves the flexibility and scalability of the application architecture.

[0012] This invention utilizes C through data envelopment analysis. 2 The R model performs a comprehensive quality and efficiency assessment of multi-source heterogeneous data sources, determines the data fusion weights based on the quality and efficiency scores, realizes data quality-driven multi-source data fusion, and effectively suppresses the negative impact of low-quality data on the fusion results.

[0013] This invention objectively determines the weight of each criterion in multi-criteria decision-making for technology selection using the entropy weight method, and performs quantitative scoring based on publicly available performance parameters of the technology. This replaces the traditional method of technology selection that relies on the subjective preferences of individual experts, thereby improving the scientificity and objectivity of technology selection. Attached Figure Description

[0014] Figure 1 A flowchart illustrating the construction method of a digital upgrade architecture for ship assembly and construction based on multi-factor comparative analysis.

[0015] Figure 2 A flowchart illustrating the process of comparative analysis of multiple elements in the construction of a digital business architecture for ship assembly.

[0016] Figure 3 A flowchart illustrating the process of establishing changes and associations for digital business elements in ship assembly and construction.

[0017] Figure 4 A flowchart illustrating the process of building a digital upgrade application architecture for ship assembly.

[0018] Figure 5 A schematic diagram of the business architecture for the digital upgrade of ship assembly.

[0019] Figure 6 A schematic diagram of the digital upgrade technology architecture for ship assembly. Detailed Implementation

[0020] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0021] like Figure 1 As shown, this embodiment provides a method for constructing a digital upgrade architecture for ship assembly and construction based on multi-factor comparative analysis, including the following steps: S1: Multi-factor Comparative Analysis of Digital Business Architecture for Ship Assembly and Construction. Taking the entire lifecycle business process of ship assembly and construction as the research object, business architecture data of typical domestic shipyards and advanced international shipyards with advanced digitalization were collected through enterprise visits, technical personnel discussions, and literature reviews. A multi-factor comparative analysis matrix was constructed, the differences in business architecture elements were calculated, and a business architecture upgrade requirement vector was generated. The multi-factor comparative analysis process of the digital business architecture for ship assembly and construction is as follows: Figure 2 As shown.

[0022] S1-1: Business Architecture Element Collection and Formal Modeling; This involves collecting data on the current business architecture of typical domestic shipyards throughout the entire lifecycle of ship assembly and construction, including supply chain management, warehousing and logistics management, 3D digital product design, sub-section manufacturing planning, sub-section welding process management, precision inspection and control, preventive equipment maintenance, full-process quality inspection, production resource scheduling, human resource allocation, yard management, and energy management. Data is collected through visits to leading domestic shipyards, discussions with technical personnel, and literature reviews. Target business architecture data is also collected through literature reviews of advanced international shipyards with advanced digitalization practices.

[0023] To ensure comparability of business structures between domestic and international shipyards, a standardized set of ship assembly and construction business domains is predefined. ,in This represents the total number of standardized business domains. This set of standardized business domains is an abstraction and summary of common business functions in shipyards both domestically and internationally, including design, procurement and supply chain, production planning, section manufacturing, final assembly, quality inspection, equipment maintenance, warehousing and logistics, human resources, yard management, and energy management. All collected current and target business architecture data are mapped to this set of standardized business domains.

[0024] Let the current business architecture elements of ship assembly and construction be: ,in This represents the total number of elements in the current business architecture. Indicates the first Each current business architecture element The value range is from 1 to A positive integer. Let the set of target business architecture elements for ship assembly and construction be... ,in The total number of elements in the target business architecture. Indicates the first One target business architecture element, The value range is from 1 to Positive integers. Business architecture elements include supply chain collaboration elements, warehousing and logistics management elements, product 3D digital design elements, segmented manufacturing planning elements, segmented welding process elements, precision inspection and control elements, equipment preventive maintenance elements, quality inspection throughout the entire process elements, production resource scheduling elements, human resource allocation elements, yard management elements, and energy management elements.

[0025] Building a formal context between current business architecture elements and standardized business domains ,in This represents the binary relationship between current business architecture elements and standardized business domains. The definition is as follows: For any and , If and only if business architecture elements Belongs to the standardized business domain The scope of responsibilities.

[0026] Based on the collected target business architecture data, adopt and build Similarly, construct a formal context between the target business architecture elements and standardized business domains. ,in This refers to the binary relationship between target business architecture elements and standardized business domains.

[0027] Based on formal background Construct the current form of the concept set For any and ,definition ,definition .but A formal concept is a concept if and only if and . This is called the extension of the formal concept. This is called the intension of a formal concept. The partial order relation between formal concepts is calculated: for any two formal concepts... and , If and only if This condition is equivalent to Based on this partial order relation, construct the current formal concept lattice. Based on formal background Adoption and Construction Similarly, construct the target formal concept lattice. .

[0028] S1-2: Construction of multi-factor comparative analysis matrix and calculation of differences; based on the current business architecture element set. Target business architecture elements set Current form concept grid and target form concept grid A multi-factor comparative analysis matrix was constructed.

[0029] Compare and analyze the differences between the current architecture and the target architecture in terms of business composition and business processes, and construct a multi-factor comparison matrix. Matrix elements Indicates the elements of the current business architecture With target business architecture elements The degree of difference between them, among which The value range is from 1 to positive integers, The value range is from 1 to Positive integers.

[0030] Difference The calculation comprehensively considers three dimensions: semantic similarity, structural similarity, and process similarity. ; in, This is the semantic similarity weight coefficient. For structural similarity weighting coefficients, These are the process similarity weight coefficients, all of which are non-negative and satisfy the following conditions: The aforementioned weighting coefficients are objectively determined using the entropy weighting method, based on the paired similarity data between the current business architecture elements and the target business architecture elements. The specific determination method is as follows: Suppose that for each pair of current elements With target elements The three similarity component values ​​were calculated: , , All The three similarity components of each element pair are organized into a data matrix. row index Iterate through all element pairs and combinations. and These are all row indices for combining all element pairs. and The value range is from 1 to positive integers, where . , , Normalize each column. ,in This provides indices for three similarity dimensions. The entropy value for each dimension is calculated. ,when Time definition Then the weighting coefficient , , This entropy weighting method assigns greater weight to similarity dimensions (smaller information entropy) with greater differences between element pairs, thereby achieving objective data-driven determination of weights.

[0031] To achieve semantic similarity based on the names and functional descriptions of business architecture elements, a semantic distance calculation method based on domain ontology is adopted. The domain ontology is a predefined knowledge base of shipbuilding business terms and relationships. The domain ontology is constructed by extracting core terms from shipbuilding industry terminology standards, shipyard business process documents, and shipbuilding engineering dictionaries, using the Protégé ontology editing tool, and storing them in the OWL standard ontology language. In this implementation, the top-level classification of the shipbuilding domain ontology includes five primary concepts: design domain ontology, production domain ontology, inspection domain ontology, logistics domain ontology, and maintenance domain ontology. Each primary concept contains several secondary concepts (e.g., the production domain ontology contains secondary concepts such as section manufacturing, final assembly, and welding processes). Synonyms between terms are defined through a terminology mapping table, and hierarchical relationships are defined through "is-a" relationships. Each concept node in the ontology has a unique identifier, and the semantic distance between concepts is measured by the number of edges along the shortest path in the directed acyclic graph of the ontology. Specifically, let business elements... The corresponding concept node in the shipbuilding field is: Business elements In the shipbuilding ontology, the corresponding concept node is c_{q}. The formula for calculating semantic similarity is: ;in Let c_{p} be the number of edges in the directed acyclic graph of the ontology between concept nodes c_{p} and c_{q}.

[0032] The structural similarity function is calculated using the following formula: ;in, Indicate business elements The corresponding object concept node in the formal concept lattice. For the current business architecture elements. Its corresponding object concept node For the current form of concept grid The extension of the middle is inclusive The formal concept corresponding to the smallest set; for the target business architecture elements Its corresponding object concept node For the target formal concept grid The extension of the middle is inclusive The formal concept corresponding to the smallest set of . Let be the centrality measure function of a node in its respective formal concept lattice, and its calculation formula is: ;in For nodes The degree of a node in a formal concept lattice, i.e., the number of edges connected to that node, where V is the set of all nodes in the current formal concept lattice. The denominator is the maximum degree of all nodes in the formal concept lattice, and the centrality value is normalized to 0. Interval.

[0033] The process similarity function is calculated using the following formula: ; in, Business elements The set of input-output relationships, containing the names of all input and output data items for this element. The input and output data item names are extracted from the business process model or data flow diagram according to the following rules: For each activity node in the business process model, the name of the data item consumed by the activity is used as the input data item, and the name of the data item produced by the activity is used as the output data item; the data item names adopt a unified naming convention in the format of "data entity.attribute"; when the same business element... When there are multiple activity nodes in the business process model Take the union of the input and output data items of all active nodes.

[0034] S1-3: Generation of Business Architecture Upgrade Requirement Vectors; Based on Multi-Factor Comparison Matrix We used an attribute reduction method based on rough set theory to extract key difference elements.

[0035] Build a decision table ,in For a collection of objects and correspond , For a set of conditional attributes and correspond , It is a decision attribute and represents the current business element. Is this a core difference? For objects... Conditional attributes The value is the degree of difference. Decision attributes The value of is determined in the following way: Calculation object Minimum difference from all target elements .like This indicates that the current element is difficult to find a similar match in the target architecture, and is a significant difference point. Decision attributes A mark of 1 indicates that an upgrade or obsolescence is needed; otherwise, a mark of 0 indicates that an upgrade or obsolescence is required. The preset difference threshold is set to 0.65 in a preferred embodiment.

[0036] For any subset of attributes Define indistinguishable relations ,in The value range is from 1 to Positive integers.

[0037] property Regarding decision attributes The positive domain is defined as ,in for of -Approximation, For object In attribute set The equivalence class below.

[0038] The goal of attribute reduction is to find the smallest subset of attributes. Make A heuristic reduction algorithm based on attribute importance is adopted, and the attributes... Importance is defined as The specific steps of the algorithm are as follows: First, initialize the reduction set. The second step is to process each unselected attribute. Calculate its relative to the current Importance The third step is to select the attribute with the highest importance. join in If there are multiple ties for the maximum value, then randomly select one; fourth step, determine... Is it equal to If the expression is equal to the given expression, stop the iteration and output the result. This is the reduction result; if it is not equal, return to step two and continue iterating. Iterate through the selection of the attribute with the highest importance until the reduction condition is met, obtaining the reduced set of key difference factors. .

[0039] Based on the set of key difference factors Generate business architecture upgrade requirement vector ,in Representing the target business architecture elements belong An upgrade is required. This indicates that no upgrade is needed.

[0040] S2: Construction of a digital business element library for ship assembly and construction. Based on the business architecture upgrade requirement vector output from S1. Current business architecture elements set Target business architecture elements set Standardized business domain set and the current business architecture concept Using fuzzy clustering and grey relational analysis as input, a digital business element library is constructed by employing a business element change method. This establishes the relationships between business elements, and the output is a digital upgrade business element library for ship assembly and construction. The process for changing and establishing relationships among digital business elements in ship assembly and construction is as follows: Figure 3 As shown.

[0041] S2-1: Construction of a digital business element library; based on business architecture upgrade requirement vectors Identify the business elements that need to be changed. (Set) With the set of target business architecture elements One-to-one correspondence by index, that is correspond Current business element library That is, the set of current business architecture elements defined in S1-2. Target business element library This refers to the set of target business architecture elements defined in S1-2. Define indicator functions As follows: For any business element ,like Furthermore, this element is reflected in the business architecture upgrade requirement vector output by S1-3. The corresponding components ,but This indicates that the element needs to be upgraded; if ,but This indicates that the element does not need to be upgraded.

[0042] according to Compared with the set of elements, business elements are divided into three categories: elements that need to be added. Elements to be deleted Elements that need to be integrated .

[0043] For elements that need to be integrated An element merging method based on fuzzy clustering is adopted. Let... Each element Using eigenvectors express, For feature dimension, The value range is from 1 to Positive integers. The feature vector is constructed as follows: Functional attribute dimension: One-hot encoding based on shipbuilding business function classification is used, with 12 business element categories as encoding bits. These 12 business element categories correspond one-to-one with the 12 business processes collected in S1-1; Process position dimension: The sequential number (an integer from 1 to 12) of the element in the entire lifecycle business process is used as the value; Data dependency dimension: The number of other elements that have direct data interaction with this element is counted. The feature vector is constructed by concatenating these three dimensions. 3D eigenvectors.

[0044] Define the fuzzy membership matrix ,in For the number of clusters, Representation elements Belongs to the The membership degree of each cluster and satisfying , The value range is from 1 to Positive integers. The objective function of fuzzy clustering is: ; in The objective function value for fuzzy clustering. This represents the total number of elements that need to be integrated. For the number of clusters, as elements Belongs to the The membership degree of each cluster, It is a fuzzy weighted index with a value of 2. as elements eigenvectors, For the first The center vectors of the clusters, It is the Euclidean norm.

[0045] By iteratively optimizing the objective function Update the membership matrix and cluster center : ;in The cluster center index has a value range from 1 to 1. The positive integers are defined as follows, and the meanings of the remaining characters are the same as those defined in the objective function above.

[0046] The meaning of each character is the same as that defined in the objective function and membership update formula above. Iterate until... Convergence occurs when the change in the objective function value between two consecutive iterations is less than a preset convergence threshold. , The value is determined based on the total number of elements to be integrated. Confirmed: When hour, ;when hour, Segmentation threshold The setting is based on: the number of elements that need to be integrated. As the value increases, the fuzzy clustering objective function... As the number of summation terms increases, the magnitude of the objective function value increases accordingly, therefore it needs to be increased accordingly. To maintain relative consistency in convergence judgments, this segmented value rule ensures that the clustering results no longer change significantly after iterations stabilize. For elements within the same cluster, if the maximum membership degree... The aforementioned elements are then merged into a new business element, and the feature vector of the new element is taken as the cluster center vector. , For a preset threshold and in a preferred embodiment, take .

[0047] After adding, deleting, and integrating elements, a business element library for the digital upgrade of ship assembly and construction is formed. .

[0048] S2-2: Establishment of element relationships based on grey relational analysis; Based on the constructed digital upgrade business element library for ship assembly and construction. Establish the relationships between business elements. Let... ,in This represents the total number of business elements after the upgrade.

[0049] For any two business elements and Grey relational analysis was used to calculate its correlation degree. and The value range is from 1 to Positive integers.

[0050] Construct a reference sequence and a comparison sequence. Let the elements... The feature vector is ,in The number of feature dimensions, The value of is the same as in S2-1 Equal. By element Reference sequence , elements For comparing sequences .

[0051] The eigenvectors are mean-normalized and dimensionlessized. ; in as elements The The value of each feature after mean normalization as elements The 1 eigenvalue, The number of feature dimensions, The feature dimension index has a value range from 1 to positive integers, The value range is from 1 to Positive integers.

[0052] Calculate the correlation coefficient. First, calculate the absolute value of all feature differences between all feature pairs, and define the two-level minimum difference. and the maximum difference between the two levels as follows: ; ; in and These are all indexes used to traverse all business elements. and The value range is from 1 to Positive integers; To iterate through the indexes of all feature dimensions, The value range is from 1 to Positive integers.

[0053] Then elements With elements In the Correlation coefficients across feature dimensions for: ; in In a preferred embodiment, the resolution coefficient is taken as... . as elements With elements In the The absolute difference across each feature dimension.

[0054] Calculate the grey relational degree: ; Grey relational degree Reflects the elements and The degree of correlation between them A larger value indicates a stronger association.

[0055] for The newly added elements are defined based on business processes, establishing their interaction relationships with existing elements. Let the newly added elements be... With existing elements The grey relational degree is ,like Then establish and The relationship between them To preset the correlation threshold, in a preferred embodiment, we take... .

[0056] for The element to be deleted is identified, and its association with all other elements is removed.

[0057] Integrate the relationships between the elements merged in S2-1. If elements... and Merged into new elements ,but With any other elements The degree of correlation is ,in The value range is from 1 to positive integers and .

[0058] S2-3: Generation of the Digital Upgrade Business Architecture Element Library; Based on the relationships between elements, a digital upgrade business element library for ship assembly and construction is generated. The element library includes: element sets. ; Element association matrix ,in as elements and The grey relational degree between them and Feature matrix ,in as elements The Each feature value.

[0059] Based on standardized business domain set The upgraded elements are then mapped back to the standardized business domain. For each upgraded element... Determine its standardized business domain. The mapping rule is: if If an element is inherited from an existing element and that element already has a standardized business domain affiliation in the original architecture, then the original standardized business domain affiliation is retained; if If a system is formed by merging multiple elements, then its subordinate standardized business domain is the union of the original elements' subordinate standardized business domains (when the union contains multiple standardized business domains, the standardized business domain containing the most original elements is taken as the final subordinate domain; if the numbers are the same, any one of them can be chosen); if For new elements, the cosine similarity between the feature vector of the new element and the center vector of each existing standardized business domain is calculated, and the new element is assigned to the standardized business domain with the highest similarity.

[0060] Specifically, for each standardized business domain Its center vector ,in For standardized business domains The set of business elements ( The definition method and the background of the upgraded form in subsequent S3-1 To maintain consistency with the formal definition, that is (For brevity, natural language will be used to describe this in S2-3 first). as elements eigenvectors. Used in S2-3 as a set of established membership relationships, the formal definition of which is given in S3-1. New elements. With standardized business domains The similarity is ,but . Build the upgraded form background ,in If and only if the standardized business domain It is an element It belongs to the standardized business domain. The value range is from 1 to positive integers, The value range is from 1 to Positive integers.

[0061] S3: Establishment of the Business Architecture for Digital Upgrade of Ship Assembly and Construction. Taking the business element library for digital upgrade of ship assembly and construction output from S2 as input, this paper establishes the business architecture layers for standardized business domains and standardized business domains based on graph neural networks and standardized business logic sorting methods based on data flow statistics, thereby constructing the business architecture for digital upgrade of ship assembly and construction.

[0062] S3-1: Construction of standardized business architecture within business domains based on graph neural networks; sorting out the standardized business domains to which business elements belong based on the digital upgrade business element library for ship assembly and construction. For each standardized business domain... Its business elements are set as follows , The value range is from 1 to Positive integers.

[0063] Build a standardized business architecture diagram within the business domain , where the set of nodes For standardized business domains Business elements, edge set The edge weights represent the grey relational degree, which indicates the relationships between elements. ,in The preset correlation threshold is the same as the value in sub-step S2-2. and The value range is from 1 to Positive integers.

[0064] A graph neural network is used to learn the representation of the business architecture within the standardized business domain and extract its feature representation.

[0065] For standardized business domains The picture Set nodes The initial feature vector is , Feature matrix of elements The Rows, dimensions are .

[0066] The message passing process in a graph neural network is as follows: ; ; in The index of the number of layers in a graph neural network and , This represents the total number of layers. The total number of layers in the graph attention network. This is a hyperparameter, determined in practical applications based on the scale of the business architecture diagram within the standardized business domain: when the number of nodes in the business architecture diagram within the standardized business domain... When less than 20 Take 2; when the number of nodes When greater than or equal to 20 and less than 50 Take 3; when the number of nodes When greater than or equal to 50 Take 4. For the first Layer message functions; For nodes The set of neighboring nodes; For the first Aggregation functions for layers.

[0067] Specifically, a graph attention network architecture is adopted: ; in It is a non-linear activation function and uses the ReLU function. For the first The layer weight matrix. Specifically, The dimension is , ( The dimension of ) is , This is the latent space dimension. In a preferred embodiment, Based on the number of nodes in the standardized business domain business architecture diagram Confirmed: When hour, ;when hour, ;when hour, Attention coefficient The calculation is as follows, where edge weights are incorporated. As prior knowledge: ; in For the first The attention parameter vector of the layer has a dimension of , This represents a vector concatenation operation. For linearly modified activation functions, The elements calculated in step S2-2 and The degree of gray correlation between them It is a very small positive number and takes the value of Used to avoid when or As the logarithm approaches zero, logarithmic operations tend towards negative infinity. It is the natural logarithm function. For traversal The index of the element in the middle. and All are nodes and index, and The value range is from 1 to Positive integers; For traversal The index of an element in the set, whose value ranges from the index of all elements in the set.

[0068] The graph attention network is trained using an unsupervised learning approach, and the training objective is an edge prediction loss function based on negative sampling, in the following form: ; The first term is the positive sample loss, which makes the final representations of node pairs connected by edges similar; the second term is the negative sample loss, which makes the final representations of node pairs not connected by edges different. This is the sigmoid function. Represents a node With nodes There is no edge between them. , , All are node indices, and their values ​​range from the set of nodes. The index of an element in the set. Negative samples are obtained by accessing the node set. Random sampling and Unconnected nodes The number of negative samples corresponding to each positive sample is a preset value, which is 5 in a preferred embodiment. When the number of positive samples... With the number of negative samples When the ratio is greater than 1:10, a weighted loss function is used, with a positive sample weight of 1 and a negative sample weight of 1. To address the issue of imbalanced positive and negative sample numbers, the Adam optimizer is used for parameter updates, with a learning rate set to 0.001. The number of training epochs depends on the total number of layers. Confirmed: When The training sessions consist of 150 rounds. The training rounds are 200 rounds. The training rounds are 250; or an early stopping strategy is adopted, which terminates training early when the loss function value on the validation set does not decrease for 20 consecutive rounds.

[0069] go through After propagation by the layered graph neural network, a standardized business domain is obtained. The final representation of each business element Standardized business domains Business architecture characteristics are represented as ,in To read the function and use average pooling, i.e. , For standardized business domains Feature vectors of business architecture The dimension of the latent space is the same as the dimension of the node representation in the final output.

[0070] S3-2: Standardized inter-domain business logic analysis based on data flow statistics; representation of business architecture characteristics based on each standardized business domain. In conjunction with the digital upgrade of ship assembly and construction, we streamlined the standardized business logic between business domains throughout the entire lifecycle of business processes.

[0071] For any two standardized business domains and Define the strength of business associations between standardized business domains : ;in and The value range is from 1 to positive integers, Cosine similarity of feature vectors for standardized business domain business architecture: ; For standardized business domains and The actual interaction intensity within a business process is objectively calculated based on existing business process data. Specifically, it is determined by extracting all business process instance data across standardized business domains from the enterprise's existing business process management system, and then statistically analyzing the standardized business domains within a preset time window (e.g., the past year). and Number of business interactions between (Unit: times), Interaction data volume (Unit: megabytes, MB) and the timeliness weight of the interaction. Among them, timeliness weight Defined as , For standardized business domains and The maximum allowed latency (in minutes) for business interactions between them. The adjustment coefficient (in a preferred embodiment, it is taken as...) This definition allows interactions with shorter latency to have a higher timeliness weight. In the calculation... Previously, for , and Perform min-max standardization on each indicator to map them uniformly to... Calculate the product after dividing by interval. The calculation formula is: ;in and Traverse all standardized business domain indexes, and The value range is from 1 to Positive integers, with the denominator being all standardized business domain pairs. The maximum value, to ensure Normalization to Interval. If there is no direct business interaction between a pair of standardized business domains, then , When business interaction data is unavailable, a method based on business process model structure analysis can be used: calculating standardized business domains. and Activity dependency path length in the business process model (Measured by the number of edges along the shortest path between active nodes), defined The shorter the path, the greater the interaction intensity. One of the two methods can be chosen based on data availability. The selection rule is: when business process instance data is available and the data volume is sufficient, the method based on... , and The statistical method; when business process instance data is unavailable or insufficient, a method based on activity dependency path length is adopted. The structural analysis method. When both types of data are available but the calculation results from the two methods differ by more than 50%, the average of the two results is taken as the mean. The value of .

[0072] Based on the strength of business associations between standardized business domains Construct a directed weighted graph of the business architecture layer between standardized business domains The node set is a standardized business domain set. edge set , To preset the correlation strength threshold, in a preferred embodiment, we take... edge weight .

[0073] S3-3: Business Architecture Integration and Optimization; Based on the standardized business architecture within and between business domains, study the logical relationship between the two, and optimize the composition and correlation of business elements.

[0074] Building a unified business architecture hypergraph The set of hypergraph nodes For all business elements; hyperedge set It includes two types of hyperedges: Hyperedge within a standardized business domain: For each standardized business domain Its set of business elements Form a superedge ; Standardized business domain hyperedges: For directed edges between standardized business domains Define hyperedge Standardized business domain Standardized business domain A set of strongly related business elements.

[0075] The optimization goal of the business architecture is to maximize the cohesion of business elements within standardized business domains and minimize the coupling of business elements between standardized business domains. ; Among them, the cohesion within the standardized business domain Defined as a standardized business domain The average value of the grey relational degree between internal business elements: ; Standardized inter-domain coupling Defined as a standardized business domain and The average value of the grey relational degree between business elements: ; Through iterative optimization The positioning of business elements in different standardized business domain architectures and standardized business domain layers of the business architecture is adjusted to optimize the composition and relationships of business elements. The optimization process employs a heuristic algorithm based on simulated annealing: in each iteration, a business element is randomly selected, and the grey relational degree between this business element and each standardized business domain to which it does not currently belong is calculated. The standardized business domain with the highest grey relational degree is selected as the target domain for a move attempt. If the grey relational degree between all selectable target domains and the business element is less than a preset move attempt threshold (in this embodiment, it is taken as...), the move attempt is considered complete. This value corresponds to a medium-low level of grey relational degree. Values ​​below this indicate no substantial relationship between the feature and the target domain, meaning moving the feature does not improve cohesion. Therefore, this threshold is set to filter out invalid move attempts and improve optimization efficiency. If a move attempt is skipped, the iteration proceeds directly to the next iteration. The objective function value after the move is calculated. .like If the move is successful, the move will be accepted; otherwise, it will be made based on probability. Accept this move, among which The current temperature. With the number of iterations from the initial temperature With attenuation coefficient Exponential decay until the termination temperature is reached. Or no movement has occurred for several consecutive iterations. In a preferred embodiment, , , .

[0076] After the optimization process is completed, the upgraded background format is updated based on the affiliation between the final business elements and the standardized business domains. This ensures that the element library output from step S2 is consistent with the business architecture of step S3. Ultimately, this forms the business architecture for the digital upgrade of ship assembly and construction.

[0077] like Figure 5 As shown, the business architecture for the digital upgrade of ship assembly and construction comprises eleven standardized business domains: design, procurement and supply chain, production planning, section manufacturing, final assembly, quality inspection, equipment maintenance, warehousing and logistics, human resources, yard management, and energy management. Each standardized business domain contains its own set of business elements, and these elements are quantitatively related through grey relational analysis. The standardized business domains are also interconnected through directed weighted relationships based on the intensity of their business interactions, forming a business architecture layer between the standardized business domains. The business architecture hypergraph uses all business elements as nodes and the hyperedges within and between standardized business domains as the set of hyperedges, comprehensively depicting the entire business architecture of the digital upgrade of ship assembly and construction.

[0078] S4: Construction of the Application Architecture for Digital Upgrade in Ship Assembly and Construction. Using the business architecture for digital upgrade in ship assembly and construction output from S3 as input, an application architecture benchmarking analysis is conducted. A modular decoupling method based on microservice architecture and a data interface design method based on event-driven architecture are adopted to construct the digital upgrade application architecture. The construction process for the application architecture for digital upgrade in ship assembly and construction is as follows: Figure 4 As shown.

[0079] S4-1: Application Architecture Benchmarking Analysis and Requirement Identification; Identifying the Functional Requirements of the Application Architecture Based on the Business Architecture of Digital Upgrade in Ship Assembly and Construction. (This is followed by a seemingly unrelated sentence about a business architecture hypergraph.) Each standardized business domain within the super-edge And hyper-border between each standardized business domain Each hyperedge is mapped to an application function module. Within a standardized business domain, the hyperedge is mapped to an application function module for a single standardized business domain. Between standardized business domains, the hyperedge is mapped to an application function module for collaboration across standardized business domains.

[0080] Through company visits, technical surveys, and conference discussions, we gained an understanding of the current status of typical domestic shipyards in terms of supply chain logistics management, product design, production information collection, and real-time data analysis. We also conducted in-depth research on the digital application architecture technology stack and design of foreign digital shipyards, analyzing the digital application innovations of leading shipyards in typical scenarios such as digital ship design, remote equipment monitoring and maintenance, and real-time workshop monitoring.

[0081] By comparing and analyzing the key differences between the current application architecture and the target digital application architecture, the architectural upgrade requirements can be clearly identified. Let the current application's functional module set be... The target application functional module set is , This represents the total number of functional modules in the current application. The total number of functional modules in the target application.

[0082] Based on the business architecture supergraph Determine the set of target application functional modules Mapping relationships with business elements. For each target functional module. The corresponding set of business elements is ,in Business architecture hypergraph output for S3 The set of nodes, and , This is the set of elements output by S2. The value range is from 1 to Positive integers.

[0083] S4-2: Application modular decoupling based on microservice architecture; based on the target application functional module set Mapping relationship with business elements A microservice architecture is used to decouple functional modules.

[0084] For each target functional module Define its service boundary If two functional modules and The corresponding sets of business elements have an intersection, i.e. Therefore, there is a data dependency between the two modules. and The value range is from 1 to Positive integers.

[0085] Define the coupling degree between functional modules: ; like Therefore, it is necessary to decouple the two modules through interface design. To preset the coupling degree threshold, in a preferred embodiment, we take... .

[0086] The functional modules are divided into the following types using the domain-driven design approach: supply chain logistics management module, product 3D digital design module, production planning and scheduling module, segmented manufacturing execution module, quality full-process control module, equipment remote operation and maintenance module, energy management module, and data integration and analysis module.

[0087] Each microservice module The architecture design includes a service interface layer, a business logic layer, a data access layer, and an infrastructure layer. The service interface layer defines the application programming interfaces (APIs) provided externally; the business logic layer implements core business functions; the data access layer is responsible for data persistence; and the infrastructure layer provides basic service support.

[0088] S4-3: Data interface design based on event-driven architecture; modular development and deployment of functional components based on microservice modules and their dependencies. Interface locations are designed according to the ship assembly and construction business process, and interface formats are determined based on business content.

[0089] An event-driven architecture is used to design data interfaces between modules. A set of event types is defined: .

[0090] For any two modules that have data interaction and and Design event channels Module As an event producer, I publish events. To the event channel Module As a consumer of the event Subscribe to and process events.

[0091] The data structure for an event is defined as follows: ; in A unique identifier for the event. For event type, Generate a timestamp for the event. For the event source module, For the business data carried by the event, It contains metadata for the event and includes priority and retry policy information.

[0092] This enables information flow between the application architecture's functional components and existing systems and data infrastructure, such as typical shipyard manufacturing execution systems and enterprise resource planning systems. For each existing system... Design adapter module Implement bidirectional conversion between event formats and existing system interface formats.

[0093] adapter The conversion function is and ,in For event payload data, For existing system interface data, A function to convert existing system data into event data. This is a function that converts event data into existing system data.

[0094] S5: Construction of Data Architecture for Digital Upgrade of Ship Assembly and Construction. Taking the step-by-step application architecture for digital upgrade of ship assembly and construction as input, a data architecture benchmarking analysis is conducted. A multi-source heterogeneous data fusion method based on data quality assessment and a hierarchical data management method based on data lake are adopted to construct the data architecture for digital upgrade.

[0095] S5-1: Data Architecture Benchmarking Analysis and Demand Identification; Based on the application architecture for digital upgrade of ship assembly and construction, identify the functional requirements of the data architecture, understand the data application needs of typical domestic shipyards through enterprise visits, technical personnel discussions and seminars, and analyze the current shipyards' data collection methods, storage paths and flow patterns for assembly and construction data such as hull design data, production process data, manufacturing quality data and supply chain data.

[0096] The study delves into the practices and achievements of advanced shipyards in digital data architecture, understanding the types of digital data architectures, technology selection, data management, and security strategies, as well as the practical experience of advanced shipyards in data integration, real-time analysis, and quality control.

[0097] Compare and analyze the differences between the current data architecture and the target data architecture in terms of data technology and management model to identify the key elements for data architecture upgrade.

[0098] Let the data source set be , This represents the total number of data sources, including hull design data sources, production process data sources, manufacturing execution data sources, quality inspection data sources, supply chain data sources, and equipment operation and maintenance data sources. Hull design data sources include 3D model data, drawing data, and bill of materials data. Production process data sources include process specification data, time quota data, and tooling list data. Manufacturing execution data sources include production schedule data, equipment status data, and personnel attendance data. Quality inspection data sources include inspection record data, accuracy data, and inspection report data. Supply chain data sources include purchase order data, bill of materials data, and inventory data. Equipment operation and maintenance data sources include equipment ledger data, maintenance record data, and fault data.

[0099] S5-2: Multi-source heterogeneous data fusion based on data quality assessment; based on data source set Data quality assessment methods are used to fuse multi-source heterogeneous data. For each data source... Define data quality metric vectors , The value range is from 1 to A positive integer, where: For completeness, it is defined as the proportion of the number of non-empty fields in the data source to the total number of fields; For accuracy, it is defined as ,in This is a set of verification samples randomly sampled from this data source. The values ​​recorded in the data source. The corresponding true value is obtained through manual verification or an authoritative third-party system. It is a very small positive number (taking values ​​of 10 ... To prevent the denominator from being zero; For consistency, it is defined as the proportion of records in this data source that completely match the key field values ​​of the same business entity in the authoritative master data source, i.e. ,in To compare the total number of records, This represents the number of records that are completely matched. For timeliness, it is defined as ,in This is the average time delay from the moment the business occurs to the moment the data becomes available from the data source, in minutes. Uniqueness is defined as the proportion of non-duplicate records in the data source to the total number of records.

[0100] C using Data Envelopment Analysis 2 The R model evaluates the overall quality of each data source. For data sources... The input metric is the cost of data collection. Data storage costs and data processing complexity Among them, data acquisition costs and data storage costs are measured in tens of thousands of yuan of the total annual cost; data processing complexity This is measured by the number of data transformation steps required to process the data source. Each step is defined as a complete logical unit from the input of an executable task (such as a Structured Query Language (SQL) script, an Extract-Transform-Load (ETL) job, or a data cleaning procedure) to its corresponding output. The number of steps is accumulated from the original interface of the data source entering the system until the data enters the standardized data layer. The output metric is the aforementioned data quality metric, namely data integrity. Data accuracy Data consistency Data timeliness and data uniqueness .

[0101] C 2 The linear programming form of the R model is: ; ; ; ;in, Number the data source currently being evaluated. To iterate through all data sources (including) in the constraints The constraint (including its own number) ensures that the efficiency ratio of all data sources does not exceed 1. For data source The overall quality efficiency score, For the first The weights of each output metric are determined by solving a linear programming problem, and their values ​​are chosen to ensure that the evaluated data source... Maximize the weighted output to weighted input ratio. This is because each data source... During the solution process, each obtains its own optimal multiplier combination, and the efficiency comparison between different data sources is a relative comparison under their respective optimal multipliers. For output index and , For the first The weights of each input metric, For input indicator index and , It is a non-Archimedean infinitesimal and takes the value in actual calculations. , The data source is indexed and the value range is from 1 to... A positive integer. By solving the above linear programming problem, we can ensure that the efficiency ratio of all data sources does not exceed 1, and that each weight is not less than... Maximize the data source under the constraints. The weighted output to weighted input ratio is used to obtain the overall quality efficiency score. and the corresponding optimal weight ( )and ( ).

[0102] like Then the data source For data envelopment analysis to be effective, the data quality must be high; if Then the data source This indicates that the data envelopment analysis is ineffective, meaning the data quality needs improvement.

[0103] For data sources that are ineffective in data envelopment analysis, identify their weak indicators and develop data quality improvement plans, thereby improving data quality through data cleaning, data standardization, and data completion operations.

[0104] A fusion strategy for multi-source heterogeneous data was constructed based on data quality assessment results. ,in For data source The data values ​​provided for the same business entity To integrate weights and This means that data sources with higher data quality have a greater weight in the fusion process.

[0105] S5-3: Construction of a layered data architecture based on a data lake; establishment of a comprehensive data governance mechanism based on the upgraded data architecture requirements of the integrated multi-source heterogeneous data and sub-step data architecture, and construction of a digital upgrade data architecture for ship assembly and construction.

[0106] A data lake architecture is adopted as the basic infrastructure for data storage and management. The data lake consists of five layers: The raw data layer stores unprocessed raw data collected from various data sources, maintaining the original format and content of the data; the standardized data layer performs format standardization, encoding unification, and data cleaning operations on the raw data to form standardized data; the thematic data layer organizes and stores standardized data according to business themes, including design themes, production themes, quality themes, supply chain themes, and equipment themes; the service data layer aggregates, calculates, and encapsulates thematic data to form directly usable service data based on application service needs; and the analytical data layer performs in-depth processing and modeling of data to form analytical datasets based on data analysis and decision support needs.

[0107] Establish unified standards for data quality and data transmission. Data quality standards include: data integrity standards requiring mandatory field coverage of ≥99%; data accuracy standards requiring data error of ≤0.1%; data consistency standards requiring cross-system data consistency of ≥98%; data timeliness standards requiring data update delay of ≤5 minutes; and data uniqueness standards requiring data duplication rate of ≤0.5%.

[0108] Establish standardized data management processes and specifications: Data collection specifications should clearly define the collection frequency, collection methods, and collection interfaces; data storage specifications should clearly define the storage format, storage period, and backup strategy; data processing specifications should clearly define the processing flow, processing tools, and processing review; and data security specifications should clearly define access permissions, encryption requirements, and audit logs.

[0109] Establish a real-time data flow pipeline to enable data flow between modules and between data sources and functional platforms: ; in Configure the data source, including the data source type, connection information, and sampling frequency. The data transformation rules include data cleaning rules, format conversion rules, and field mapping rules. Configure the data target, including the target storage location, target format, and write strategy. It is a data flow scheduling strategy that includes triggering conditions, execution frequency, and retry strategy.

[0110] S6: Construction of the Digital Upgrade Technology Architecture for Ship Assembly and Construction. Taking the digital upgrade data architecture for ship assembly and construction output from S5 as input, and combining the business architecture output from S3 and the application architecture requirements output from S4, a technical architecture benchmarking analysis is conducted. A multi-criteria decision-making method based on technology applicability assessment is adopted to select technologies and construct the digital upgrade technology architecture.

[0111] S6-1: Technical Architecture Benchmarking Analysis and Technical Requirements Identification; Based on data architecture requirements, application architecture requirements, and business architecture requirements, through enterprise visits and discussions with technical personnel, the technical components of digital construction in typical domestic shipyards are sorted out, and the existing technical architecture of digital construction of ship assembly and construction is clarified.

[0112] This study investigates the composition of the digital construction technology architecture of advanced foreign shipyards through literature review and industry benchmark analysis, analyzing the compatibility of Industrial Internet of Things (IIoT) technology and 5G fifth-generation mobile communication networks with typical business scenarios in ship assembly and construction. By benchmarking the digital construction technology architecture of typical domestic and international shipyards, a technical architecture goal for the digital upgrade of ship assembly and construction is formulated. The technical requirements for digital upgrades at different stages of ship assembly and construction are broken down as follows: Design stage requirements include 3D collaborative design platform requirements, product data management requirements, and design simulation verification requirements; Production stage requirements include advanced planning and scheduling system requirements, manufacturing execution system requirements, real-time workshop monitoring requirements, and intelligent production scheduling requirements; Management stage requirements include enterprise resource planning system requirements, supply chain management requirements, and project management requirements; Service stage requirements include remote equipment operation and maintenance requirements, digital twin requirements, and predictive maintenance requirements; Basic support stage requirements include IIoT platform requirements, 5G fifth-generation mobile communication network requirements, cloud computing infrastructure requirements, and data platform requirements.

[0113] S6-2: Technology selection based on multi-criteria decision-making; based on the technology requirements of each stage and combined with the development trend of digital technology, select the next-generation information technology for typical scenarios in different stages of final assembly and construction, such as ship design, ship production, ship management, ship service and basic support.

[0114] A multi-criteria decision-making model for technology selection is established by analyzing the applicability, scalability, integration, and economic efficiency of relevant technologies. Let the set of candidate technologies be denoted as . , The total number of candidate technologies is and the set of evaluation criteria is . ,in For the first One evaluation criterion, The value range is from 1 to 4. Applicability refers to the degree to which the technology matches the ship assembly and construction scenario. Scalability refers to the ability of technology to support future business expansion. Integration refers to the difficulty of integrating the technology with existing systems. Economic efficiency refers to the implementation and maintenance costs of the technology.

[0115] Constructing a decision matrix ,in For technology In the guidelines The rating below, The value range is from 1 to positive integers, The value range is a positive integer from 1 to 4. Instead of expert scoring, it is objectively calculated based on publicly available performance parameters and matching indices. Applicability score Computing technology Functional characteristics and functional requirements of typical ship assembly and construction scenarios The degree of matching. Let... Include Each functional requirement (such as real-time data acquisition, equipment control, data analysis, etc.) requires... Judgment technology Does this feature support the function? If it does, the score is 1; otherwise, it is 0. For indicator functions, when technology Supported functional requirements The value is 1 if the condition is met, and 0 otherwise. This refers to the proportion of technically supported functionalities out of all functional requirements. (Scalability score) Number of publicly available extension interfaces and modularity (Measured by the number of functional units that can be independently packaged, deployed, and upgraded, supported by the technology, such as the number of microservice instances or plugins. This value can be obtained from the architecture design document or official technical white paper of the technology product.) This means standardizing to the maximum value. (Integration score) Number of standardized protocols based on technical support (Such as OPCUA, MQTT, RESTful, etc.) computation. ,in The maximum number of protocols among all candidate technologies. Economic score. Implementation costs based on technology and annual maintenance costs calculate. The lower the cost, the higher the score.

[0116] If the aforementioned quantitative data is unavailable, relevant numerical indicators can be obtained by consulting publicly available documents such as technical white papers and industry benchmark reports. All data sources must be specified in the technical architecture plan.

[0117] Normalize the decision matrix .

[0118] The entropy weight method is used to determine the weights of each criterion. The entropy value is ,in In particular, when When, define .

[0119] Guidelines The weight is ,in The value of is a positive integer from 1 to 4.

[0120] technology The overall score is .

[0121] The technology with the highest overall score was selected as the technology selection solution for this field. The technology selection for the digital upgrade of ship assembly and construction was completed by considering the resource constraints of the ship assembly and construction business scenario, and a digital technology architecture element library for ship assembly and construction was constructed.

[0122] S6-3: Digital Upgrade Technology Architecture Integration and Construction; Based on the completed technology selection, and combined with data architecture requirements, application architecture requirements, and business architecture requirements, research the deep integration methods of industrial Internet of Things technology and 5G fifth-generation mobile communication network with the existing technology system of ship assembly and construction, and form the core components of the digital upgrade technology architecture for ship assembly and construction.

[0123] The technical architecture is divided into five layers: the equipment layer includes various intelligent production equipment, sensors, actuators, RFID tags, and smart terminals; the network layer includes 5G private mobile communication networks, industrial Ethernet, WiFi 6 wireless networks, and IoT gateway communication infrastructure; the data layer includes a five-layer data lake architecture, real-time data flow pipelines, and a data governance platform; the application layer includes a collection of microservice modules, including supply chain logistics management modules, product 3D digital design modules, production planning and scheduling modules, segmented manufacturing execution modules, quality control throughout the entire process modules, remote equipment operation and maintenance modules, energy management modules, and data integration and analysis modules; and the presentation layer includes various user interfaces, various visual dashboards, various mobile applications, and a digital twin visualization platform.

[0124] like Figure 6 As shown, the digital upgrade technology architecture for ship assembly and construction is divided into five layers from bottom to top: equipment layer, network layer, data layer, application layer, and presentation layer. The equipment layer includes various intelligent production equipment, sensors, actuators, RFID tags, and intelligent terminals, responsible for data acquisition and execution control at the assembly and construction site. The network layer includes a 5G fifth-generation mobile communication private network, industrial Ethernet, WiFi 6 wireless network, and IoT gateways, responsible for providing a highly reliable, low-latency data transmission channel between the equipment layer and the data layer. The data layer includes a five-layer data lake architecture, real-time data flow pipelines, and a data governance platform, responsible for the storage, processing, and analysis of multi-source heterogeneous data. The application layer includes eight microservice modules: supply chain logistics management, product 3D digital design, production planning and scheduling, segmented manufacturing execution, quality control throughout the entire process, remote equipment maintenance, energy management, and data integration and analysis, responsible for realizing digital business functions throughout the entire lifecycle of ship assembly and construction. The presentation layer includes various user interfaces, visual dashboards, mobile applications, and a digital twin visualization platform, responsible for providing users with intuitive interactive and decision support interfaces.

[0125] Clarify the technical logical relationships between different levels: The device layer transmits the collected data to the network layer via 5G fifth-generation mobile communication network and industrial Ethernet; the network layer aggregates the data and transmits it to the data layer via IoT gateway; the data layer stores, processes and analyzes the data and provides services to the application layer via application programming interface gateway; the application layer implements business functions through microservice architecture and provides an interactive interface to users through presentation layer.

[0126] The data flow relationship between layers is ,in For the first Layer data, For the first Layer configuration parameters, For the first layer to the first Layer data transformation functions, The value range is a positive integer from 1 to 4, and corresponds to the device layer to the network layer, the network layer to the data layer, the data layer to the application layer, and the application layer to the presentation layer, respectively.

[0127] By combining the analysis of the entire life cycle business process of the shipbuilding industry to support the construction methods at all levels, a digital technology architecture for ship assembly and construction is constructed.

Claims

1. A method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison and analysis, characterized in that: Includes the following steps: S1: Obtain current business architecture data and target business architecture data, define a standardized set of business domains, construct the current form background, current form concept grid, target form background and target form concept grid, and generate a business architecture upgrade requirement vector; S2: Based on the business architecture upgrade requirement vector, classify and merge business elements, establish the relationship between the merged business elements, and generate a business element library for the digital upgrade of ship assembly and construction. S3: Based on the business element library for the digital upgrade of ship assembly and construction, a business architecture diagram is constructed for each standardized business domain in the standardized business domain set, and the business architecture feature representation is extracted. A business architecture layer between standardized business domains is constructed, the cohesion of business elements within standardized business domains and the coupling of business elements between standardized business domains are optimized, and a business architecture for the digital upgrade of ship assembly and construction is generated. The business architecture for the digital upgrade of ship assembly and construction includes a business architecture hypergraph. S4: Map the hyperedges in the business architecture hypergraph to application function modules, decouple the application function modules and design the data interfaces between each module to generate the application architecture for digital upgrade of ship assembly and construction. S5: Based on the application architecture for digital upgrade of ship assembly and construction, data quality assessment is performed on multi-source heterogeneous data sources. Based on the data quality assessment results, multi-source heterogeneous data is fused. The fused data is managed using a layered architecture to generate a data architecture for digital upgrade of ship assembly and construction. S6: Based on the data architecture for digital upgrade of ship assembly and construction, and combined with the requirements of the business architecture and application architecture for digital upgrade of ship assembly and construction, identify technical requirements and select technologies to generate the technical architecture for digital upgrade of ship assembly and construction.

2. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific methods for constructing the current formal concept lattice and the target formal concept lattice in S1 are as follows: Construct a set of current formal concepts based on the current formal context. For any subset of objects and subset of attributes in the current formal context, define the set of common attributes of the object subset and the set of common objects of the attribute subset. Determine a formal concept as the condition pair that satisfies the condition that the set of common attributes of the object subset equals the attribute subset and the set of common objects of the attribute subset equals the object subset. Use the extension of the formal concept as the object subset and the intension of the formal concept as the attribute subset. Calculate the partial order relation between formal concepts and construct the current formal concept lattice based on the partial order relation. Construct the target formal concept lattice based on the target formal context using the same method.

3. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 2, characterized in that, The specific method for generating the business architecture upgrade requirement vector in S1 is as follows: The semantic similarity, structural similarity, and process similarity between the current business architecture elements and the target business architecture elements are calculated. Semantic similarity is calculated based on the semantic distance between terms in the shipbuilding domain ontology. Structural similarity is calculated based on the difference between the centrality measure of the corresponding node of the current business architecture element in the current formal concept lattice and the centrality measure of the corresponding node of the target business architecture element in the target formal concept lattice. Process similarity is calculated based on the ratio of the intersection to the union of the input-output relationship set of the current business architecture element and the input-output relationship set of the target business architecture element. The weighted sum of semantic similarity, structural similarity, and process similarity is used as the comprehensive similarity. The difference is calculated by subtracting the comprehensive similarity from one. A multi-element comparison matrix is ​​constructed with the current business architecture elements as rows and the target business architecture elements as columns. Key difference elements are extracted from the multi-element comparison matrix using the attribute reduction method of rough set theory. The business architecture upgrade requirement vector is generated using the key difference elements.

4. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 3, characterized in that, The weights of semantic similarity, structural similarity, and process similarity are determined using the entropy weight method. Specifically, the pairwise similarity components of all current business architecture elements and target business architecture elements are organized into a data matrix. Each column of the data matrix is ​​normalized, and the entropy value of each dimension is calculated. The normalized reciprocal of the entropy value of each dimension is used as the weight of each dimension.

5. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific method for classifying and merging business elements in S2 is as follows: Based on the business architecture upgrade requirement vector, elements that exist in the target business architecture element set but not in the current business architecture element set and need to be upgraded are designated as elements to be added; elements that exist in the current business architecture element set but not in the target business architecture element set are designated as elements to be deleted; and elements that exist in both the current and target business architecture element sets and need to be upgraded are designated as elements to be integrated. Each element to be integrated is represented as a feature vector containing functional attributes, process location, and data dependency dimensions. A fuzzy membership matrix is ​​constructed. After weighting the membership degree with a fuzzy weighting exponent, the weighted sum of the squared Euclidean distances between each element's feature vector and each cluster center vector is calculated as the fuzzy clustering objective function. The membership matrix and cluster centers are iteratively updated until the fuzzy clustering objective function converges. For elements in the same cluster, elements with a maximum membership degree exceeding a preset membership degree threshold are merged into a new business element, and the feature vector of the new element is taken from the cluster center vector.

6. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 5, characterized in that, The specific method for establishing the correlation between merged business elements in S2 is as follows: using the feature vector of each merged business element as a sequence, the feature vector is averaged and dimensionless. The minimum absolute value of all feature differences between all pairs of business elements is calculated as the two-level minimum difference, and the maximum absolute value of all feature differences between all pairs of business elements is calculated as the two-level maximum difference. For any two business elements in any feature dimension, the correlation coefficient of that dimension is obtained by multiplying the two-level minimum difference and the resolution coefficient by the sum of the two-level maximum differences and dividing by the sum of the absolute difference and the resolution coefficient of that feature dimension. The average value of the correlation coefficients of all feature dimensions is used as the gray correlation degree between the two business elements. The element correlation matrix is ​​constructed using all gray correlation degrees.

7. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific method for constructing a business architecture graph and extracting business architecture feature representations in S3 is as follows: A business architecture graph within a standardized business domain is constructed using business elements as nodes and the relationships between element pairs whose gray relational degree exceeds a preset relational degree threshold as edges. The feature vectors of the business elements are used as the initial feature vectors of the nodes. A graph attention network is used for multi-layer message passing. In each layer, the attention coefficients of neighboring nodes to the central node are calculated, and the edge weights are incorporated into the attention coefficients as bias terms. The features of neighboring nodes are weighted and aggregated using the attention coefficients as weights, and then transformed by a non-linear activation function to obtain the node features output by that layer. After multi-layer propagation, the final feature representations of each node are obtained. The average pooling of the final feature representations of each node is used as the business architecture feature representation of the standardized business domain.

8. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific method for constructing the standardized business architecture layer between business domains in S3 is as follows: Extract business process instance data across standardized business domains from the business process management system, count the number of business interactions, the amount of interaction data, and the interaction timeliness weight between any two standardized business domains within a preset time window, and use the product of the number of business interactions, the amount of interaction data, and the interaction timeliness weight as the original interaction strength between the two standardized business domains. Normalize the original interaction strength by using the maximum original interaction strength among all pairs of standardized business domains as the denominator to obtain the interaction strength between standardized business domains. When business interaction data cannot be obtained, use the reciprocal of the activity dependency path length of the two standardized business domains in the business process model plus one as the interaction strength between standardized business domains. Construct a directed weighted graph of the standardized business architecture layer between standardized business domains based on the product of the interaction strength between standardized business domains and the cosine similarity of the feature vectors of the business architecture of the standardized business domains.

9. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific method for optimizing the cohesion of business elements within standardized business domains and the coupling of business elements between standardized business domains in S3 is as follows: taking the membership relationship between the current business element and the standardized business domain as the initial state, the average gray relational degree between business elements within all standardized business domains is used as the cohesion, the average gray relational degree between business elements of all standardized business domain pairs with business interactions is used as the coupling, and the sum of the cohesion of all standardized business domains minus the sum of the coupling between all standardized business domains is used as the optimization objective function. In each iteration, a business element is randomly selected and its movement from its current standardized business domain to another standardized business domain is attempted. The optimization objective function value after the movement is calculated. If the optimization objective function value after the movement is greater than the optimization objective function value before the movement, the movement is accepted. Otherwise, the movement is accepted based on the probability calculated based on the current temperature. The temperature decreases exponentially with the number of iterations according to the decay coefficient until the termination temperature is reached or the iteration converges.

10. The method for constructing a digital upgrade architecture for ship assembly and construction based on element comparison analysis according to claim 1, characterized in that, The specific method for data quality assessment and fusion of multi-source heterogeneous data in S5 is as follows: For each data source, define a data quality indicator vector including completeness, accuracy, consistency, timeliness, and uniqueness; and use the C++ algorithm of Data Envelopment Analysis. 2 The R model evaluates the overall quality efficiency of each data source, using the data acquisition cost, data storage cost, and data processing complexity of each data source as input indicators and the components of the data quality indicator vector as output indicators. It obtains the overall quality efficiency score of each data source by solving a linear programming problem, and uses the overall quality efficiency score as the quantitative evaluation result of data quality. The proportion of the overall quality efficiency score of each data source to the sum of the overall quality efficiency scores of all data sources is used as the fusion weight. The data values ​​of the same business entity provided by each data source are weighted and summed using the fusion weight to obtain the fused data.

Images